v0.13.0

VoiceBlender/voiceblenderv0.13.0Sep 4, 2026by csiwek

AI Summary

This release introduces configurable SIP proxy settings, adds padding to mono recordings, and integrates Speechmatics for speech-to-text transcription. It also includes various bug fixes regarding TLS parsing, RTP port allocation, and audio pacing, alongside new REST endpoints for managing custom leg data.

Key Highlights

Configurable SIP proxy for outbound requests
Speechmatics STT integration for room-wide transcription
New REST and VSI endpoints for custom leg data management
Fixes for TLS URI parsing and RTP port recycling
Room transcribes and records every audio source

New Features

Configurable SIP proxy
Mono recording padding
Speechmatics STT integration
Room-wide audio transcription and recording
REST and VSI endpoints for custom leg data (CRUD)
WS ingress playout lead and mixer-clocked source

Full Release Notes

## What's Changed
* Configurable SIP proxy for outbound requests by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/139
* Add padding to mono recordings by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/140
* Fix Contact URI on outgoing register by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/141
* Fix tls uri parsing by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/150
* siprec: check recording metadata against the SDP it arrived with by @agranig in https://github.com/VoiceBlender/voiceblender/pull/144
* mixer: assemble whole frames from a source of any packetisation by @agranig in https://github.com/VoiceBlender/voiceblender/pull/145
* room: transcribe and record every audio source, not just leg participants by @agranig in https://github.com/VoiceBlender/voiceblender/pull/147
* feat(wsmedia,mixer): WS ingress playout lead and mixer-clocked source… by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/151
* Makre Recording test durable by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/152
* Speechmatics STT by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/135
* fix: rtp port recycling, allocator by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/153
* fix: stop double-pacing the leg-agent audio path by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/157
* Only route Meta inbound to WhatsApp when media is still DTLS-encrypted by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/158
* Add custom data to a leg by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/159
* REST and VSI endpoints to modify custom_data for legs (CRUD) by @csiwek in https://github.com/VoiceBlender/voiceblender/pull/160


**Full Changelog**: https://github.com/VoiceBlender/voiceblender/compare/v0.12.2...v0.13.0